Inflammatory and nutritional indices as predictors of poor prognosis in pyogenic liver abscesses: a retrospective cohort study
Bihong Sun, Zhelong Liu, Shaohua Liu, Tian Yu, Li LiuObjective
This retrospective study of 458 pyogenic liver abscesses (PLA) patients explored the link between inflammation/nutrition-based indices and poor prognosis in PLA, aiming to enhance the prediction of poor prognosis such as mortality, septic shock, or intensive care unit (ICU) admission.
Methods
Patient data, including demographics, comorbidities, clinical presentations, laboratory results, and treatment outcomes, were extracted from medical records. Multivariable logistic regression and receiver operating characteristic (ROC) curve analyses were performed to identify independent predictors and evaluate their discriminative power. A simplified joint prediction model was developed based on rigorous variable selection criteria, and its performance was assessed using internal validation.
Results
Patients with poor prognosis ( n = 33) exhibited lower lymphocyte counts, hemoglobin, platelets, and albumin, alongside elevated inflammation markers. Multivariable analysis identified the neutrophil-to-albumin ratio (NAR) (odds ratio (OR) 6.87 per standard deviation (SD), 95% confidence interval (CI) [2.13–22.17], p = 0.001), prognostic nutritional index (PNI) (OR 0.99 per SD, 95% CI [0.894–0.997], p = 0.038), and controlling nutritional status (CONUT) score (OR 1.37 per SD, 95% CI [1.164–1.602], p < 0.001) as independent predictors. ROC analysis showed these indices had good discriminative ability (PNI: area under the curve (AUC) = 0.757, p < 0.001; PLR: AUC = 0.704, p < 0.001; CONUT: AUC = 0.803, p < 0.001; PLT: AUC = 0.740, p < 0.001). Notably, a joint prediction model combining platelet count (PLT) and CONUT achieved superior prediction with an AUC of 0.822.
Conclusion
This investigation reveals that easily obtainable inflammatory and nutritional indices (NAR, PNI, CONUT and PLT) serve as reliable predictors of poor prognosis in PLA. A combined model integrating PLT and CONUT demonstrates promising discriminative performance, enabling early risk stratification.